News for hackers. Stuff that matters.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by chrsig on Monday October 16, 2023 @03:12PM (#37904735)
Perhaps not useful to the conversation, but I really wish that whomever coined the behavior as a 'hallucination' had consulted a dictionary first.

It's delusional, not hallucinated.

Delusions are the irrational holdings of false belief, especially after contrary evidence has been provided.

Hallucinations are false sensations or perceptions of things that do not exist.

May some influential ML person read this and start to correct the vocabulary in the field :)

Re:Hallucineted CVE against Curl: someone asked Bard to find... by 65a on Monday October 16, 2023 @03:18PM (#37904838)
Confabulation seems better aligned to neuropsychology, as far as I can tell: https://en.wikipedia.org/wiki/Confabulation
Re:Hallucineted CVE against Curl: someone asked Bard to find... by garba_dlm on Monday October 16, 2023 @03:44PM (#37905324)
cool, the scientific name for studying gaslighting!
Re:Hallucineted CVE against Curl: someone asked Bard to find... by groby_b on Monday October 16, 2023 @03:58PM (#37905570)
Gaslighting and confabulation are very different things.

Gaslighting are deliberate lies with the intent of creating self-doubt in the targeted person. Confabulation is creating falsehoods without an intent to deceive.

When we're discussing naming, it might be a good idea not to throw more misleading names onto the bonfire.

Re:Hallucineted CVE against Curl: someone asked Bard to find... by garba_dlm on Monday October 16, 2023 @07:23PM (#37908051)
my point is they're related. but i didn't explain why

and you're mistaken on the intent of gaslighting. it's intended to control

'confabulation' reflects on how that's sort of control is even possible

Re:Hallucineted CVE against Curl: someone asked Bard to find... by lazide on Monday October 16, 2023 @08:26PM (#37908576)
Gaslighting is also usually associated with disorders involving strong delusional behavior. Delusions are maladaptive protective behavior (a false worldview that avoids actual ‘dangerous’ thoughts or information), and when challenged in a threatening way, particularly dangerous folks often gaslight the threat. It’s easy and natural for them to do, because they already have all the tools necessary to maintain the original delusion.

It’s the ‘my world view will be unchallenged or I will destroy yours’ reaction.

NPD being a very common example. Certainly not the only one though!

Re:Hallucineted CVE against Curl: someone asked Bard to find... by nikanj on Monday October 16, 2023 @03:58PM (#37905588)
That one is not part of everyone’s vocabulary
Re:Hallucineted CVE against Curl: someone asked Bard to find... by baq on Monday October 16, 2023 @04:02PM (#37905657)
Not too late still, though an uphill battle
Re:Hallucineted CVE against Curl: someone asked Bard to find... by inopinatus on Monday October 16, 2023 @05:25PM (#37906883)
If we’re going to play this game then my imaginary friend Apophenia wants to join in.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by oxygen_crisis on Monday October 16, 2023 @08:29PM (#37908613)
That's even better, then, to address the issue of laypeople misinterpreting a distinctive problem according to familiar, overloaded definitions of the word used to refer to it.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by SkyPuncher on Monday October 16, 2023 @03:39PM (#37905216)
I think it’s appropriate.

Delusion tends to describe a state of being, in the form of delusional. Hallucinations tend to be used to describe an instance or finite event.

Broadly, LLMs are not delusions, but they do perceive false information.

Re:Hallucineted CVE against Curl: someone asked Bard to find... by Brian_K_White on Monday October 16, 2023 @03:19PM (#37904856)
The llm has neither of these, so neither is more correct or incorrect than the other.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by giantrobot on Monday October 16, 2023 @03:17PM (#37904832)
I prefer the term confabulation. To the AI the made up thing isn't necessarily irrational. It's in fact very rational, simply incorrect.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by asadotzler on Monday October 16, 2023 @05:34PM (#37906977)
aka bullshit. it is a bullshitter or a bullshit artist, virtually synonymous with confabulator.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by user_7832 on Monday October 16, 2023 @03:44PM (#37905315)
I propose using delirium/delirious to describe the software.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by inopinatus on Monday October 16, 2023 @04:22PM (#37905978)
LLMs don’t hold beliefs. Believing otherwise is itself a delusion.

In addition, the headline posted here doesn’t even say hallucinated, so that is also a hallucination. It says hallucineted. As portmanteaux go, that ain’t bad. I rather like the sense of referring to LLMs as hallucinets.

Re:Hallucineted CVE against Curl: someone asked Bard to find... by bregma on Monday October 16, 2023 @04:48PM (#37906374)
I think the right word is "bullshit". LLMs are neither delusional nor hallucinating since they have no beliefs or sensory input. The just generate loads of fertilizer and a lot of people like to spread it around.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by asadotzler on Monday October 16, 2023 @05:32PM (#37906960)
This is the correct answer. It's not a hallucination. It's goal is to create something that seems like the truth despite the fact that it has no idea if it's actually being truthful. If a human were doing this we'd call them a bullshitter or of they were good at it, maybe even a bullshit artist.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by wlonkly on Tuesday October 17, 2023 @09:41AM (#37914901)
I've been calling it bullshit too, because the thing about bullshitting is that the truth is irrelevant to a good story.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by jazzyjackson on Monday October 16, 2023 @05:02PM (#37906582)
IMHO it's fine to have a certain jargon within the context of "things neural nets do" and comes from the days of Deep Dream, when image classifiers were run in reverse and introduced the public to computer-generated-images that were quite psychedelic in nature. It's seeing things that aren't there.
Re:Hallucineted CVE against Curl: someone asked Bard to find... by dylan604 on Monday October 16, 2023 @06:39PM (#37907620)
The phrase "you must be trippin'!" is commonly used by some when they say something completely nonsensical. I can easily see where how/why hallucinating was chosen.

It's clearly meant to poke fun of the system. If you think people are going to NOT use words in jest while making fun of something, perhaps you could use a little less starch in your clothing.

Re:Hallucineted CVE against Curl: someone asked Bard to find... by SrslyJosh on Monday October 16, 2023 @07:37PM (#37908148)
LLMs do not have beliefs, so "delusion" is no better than "hallucination". As statistical models of texts, LLMs do not deal in facts, beliefs, logic, etc., so anthropomorphizing them is counter-productive.

An LLM is doing the exact same thing when it generates "correct" text that it's doing when it generates "incorrect" text: repeatedly choosing the most likely next token based on a sequence of input and the weights it learned from training data. The meaning of the tokens is irrelevant to the process. This is why you cannot trust LLM output.